Commodity futures option valuation – An ensemble model

This study offers an in-depth examination of futures options valuation, a multifaceted issue due to its reliance on both the underlying futures contract and the commodity's spot price. We introduce a novel Clustering-based HAR-Ensemble model (CluEnsem) that fuses three key elements: a modified Heterogeneous Autoregressive (HAR) model, a two-layer stacking-based ensemble machine learning model equipped with a meta- learning mechanism, and a clustering mechanism. This model is designed to navigate the complex term structures and fluctuating volatility inherent in futures options. We validate our methodology using options underpinned by four key futures contracts: S&P 500 index futures, Henry Hub Natural Gas, Soybeans, and Gold, achieving exceptional performance across all assets. This study significantly advances futures options valuation literature by modeling the intricacies of implied volatility across varying maturities and proposing a clustering-based ensemble model within a single framework. Our methodology surpasses other established models, thus proving its effectiveness.